Skip to Main Content
The time series prediction model based on neural network can perfectly reflect the trend of development of nonlinear system, but the training speed for neural network is very slow, therefore, it is easily prone to local extremum. So we come up with a learning algorithm combining genetic algorithm and BP algorithm for the training of BP neural network, to realize optimization of network structure. We have built a prediction model for aircraft engine wear based o this type of algorithm. Comparisons have been made between the results from this prediction model and those from multiple linear regression method. The final test results indicate that genetic algorithm-based BP neural network is superior to BP algorithm and multiple linear regression method, bringing about much better forecasting results.